See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283334976

Electron-Muon Ranger: performance in the MICE Muon Beam

Article in Journal of Instrumentation · October 2015

DOI: 10.1088/1748-0221/10/12/P12012 · Source: arXiv

CITATIONS READS

25 248

155 authors, including:

Dianna Adams A. Alekou

California State University, East Bay CERN
177 PUBLICATIONS 9,962 CITATIONS 47 PUBLICATIONS 2,793 CITATIONS

SEE PROFILE SEE PROFILE

Marco Apollonio Ruslan Asfandiyarov

Diamond Light Source University of Geneva
321 PUBLICATIONS 13,300 CITATIONS 188 PUBLICATIONS 12,166 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by François Drielsma on 11 November 2015.

The user has requested enhancement of the downloaded file.

 RAL-P-2015-008 Muon Ionization Cooling Experiment November 6, 2015

Electron-Muon Ranger:
performance in the MICE Muon Beam
The MICE collaboration†

The Muon Ionization Cooling Experiment (MICE) will perform a detailed study of ionization
cooling to evaluate the feasibility of the technique. To carry out this program, MICE requires
an efficient particle-identification (PID) system to identify muons. The Electron-Muon Ranger
(EMR) is a fully-active tracking-calorimeter that forms part of the PID system and tags muons that
traverse the cooling channel without decaying. The detector is capable of identifying electrons
arXiv:1510.08306v2 [physics.ins-det] 3 Nov 2015

with an efficiency of 98.6%, providing a purity for the MICE beam that exceeds 99.8%. The
EMR also proved to be a powerful tool for the reconstruction of muon momenta in the range
100–280 MeV/c.

1 Introduction

Intense muon sources are required for a future Neutrino Factory or Muon Collider [1, 2]. At production, muons
occupy a large phase-space volume (emittance), which makes them difficult to accelerate and store. Therefore,
the emittance of the muon beams must be reduced, i.e the muons must be “cooled”, to maximise the muon flux
delivered to the accelerator. Conventional cooling techniques applied to muon beams [3] would leave too few
muons to be accelerated since the muon lifetime is short (τµ ∼ 2.2 µs). Simulations indicate that the ionization-
cooling effect builds quickly enough to deliver the flux and emittance required by the Neutrino Factory and
the Muon Collider [4, 5]. The MICE collaboration will study ionization cooling in detail to demonstrate the
feasibility of the technique [6].
Ionization cooling proceeds by passing a beam of muons through a low-Z material [7]. The beam loses
energy by ionizing the material, reducing its total momentum. Longitudinal momentum is restored by acceler-
ating cavities. The net effect is to reduce the divergence of the beam and the transverse phase-space the beam
occupies. The rate of change of the normalised 2D emittance may be approximated by [8]:

dεN εN dEµ β⊥ (0.014)2

'− 2 + 3 ; (1)
ds β Eµ ds 2β Eµ mµ X0

where β = v/c, Eµ , mµ are the muon velocity, energy and mass respectively. The rate of change of emittance
depends on the properties of the absorber and the beam. Cooling is large when the initial emittance of the
beam, εN , and stopping power of the absorber, hdEµ /dsi, are large. The effect of heating by multiple Coulomb
scattering is reduced if the radiation length of the absorber, X0 , is large and the transverse betatron function,
β⊥ , of the beam at the absorber is small. Optimum cooling is achieved with low-Z absorbers, such as liquid
hydrogen or lithium hydride, and with solenoidal beam-focussing.
The muon beams at the front-end of a Neutrino Factory or Muon Collider are expected to be similar, with a
large transverse normalised emittance of εN ≈ 12–20 π mm-rad and a momentum spread of ∼ 20 MeV/c. The
emittance must be reduced to 2–5 π mm-rad for the Neutrino Factory, with further reduction to 0.008 π mm-rad
required for a Muon Collider [9]. The Muon Ionization Cooling Experiment (MICE) [10] collaboration intends
to demonstrate the feasibility of an ionization-cooling cell suitable for cooling muon beams at a Neutrino
†
Authors are listed at the end of this paper.
Factory. An accurate measurement of the degree of cooling achieved depends on selecting a pure sample
of muons by rejecting their parent particles or decay products. The Electron-Muon Ranger (EMR) [11] is a
key component of the particle identification system and its performance in the MICE Muon Beam (MMB) is
presented here.

1.1 The Muon Ionization Cooling Experiment

Since energy loss by ionization and multiple Coulomb scattering are momentum dependent, the ionization-
cooling effect is momentum dependent. MICE will measure the performance of its cooling cell for muon
beams with central momenta in the range 140–240 MeV/c and a momentum spread of ∼ 20 MeV/c, as would
be the case in a Neutrino Factory. This is achieved using a “super focus-focus” lattice cell [12] which is capable
of producing a range of β⊥ at the position of the absorber.
A schematic of the MICE Muon Beam (MMB) and the MICE experiment is shown in figure 1 and described
in detail in [13, 14]. The MMB operates on the ISIS proton synchrotron [15] at the Rutherford Appleton
Laboratory. A titanium target [16] samples the ISIS proton beam, producing pions. The pions are transported
by the upstream quadrupoles, Q1–3, and are momentum-selected at the first dipole, D1. The high field present
in the Decay Solenoid (DS) increases the pion path-length with the result that the bulk of pion decays occur
as the beam passes through the DS. The second dipole, D2, is used to momentum-select a ‘muon’ beam with
high purity or a ‘calibration’ beam containing a mix of muons, pions and electrons. The resultant beam is
transported through two quadrupole triplets, Q4–6 and Q7–9, time-of-flight counters [17], TOF0 and TOF1,
and Cherenkov detectors, Ckov [18], to the cooling cell. The beam emittance is inflated as it passes through
the variable-thickness brass and tungsten ‘diffuser’ and is measured in the upstream spectrometer solenoid
using a scintillating-fibre tracker. The beam then passes through low-Z absorbers and RF cavities prior to
being remeasured in the downstream tracker. Upon exiting the cooling cell the beam is incident upon the final
time-of-flight detector, TOF2 [19], a pre-shower detector, KL [18], and the Electron-Muon Ranger, EMR [20].
The particle identification system consists of the TOF0, TOF1 and TOF2, Ckov, KL and EMR detectors,
which can conceptually be split in two; TOF0–1 and Ckov identifying particle species prior to the cooling
cell and TOF2, KL and EMR identifying species after the cooling cell. In combination with time-of-flight
information, the EMR is used to distinguish between muons that have successfully traversed the full cooling
cell from those that have decayed en-route. This aids in the measurement of beam transmission through the
cooling cell, as well as reducing the uncertainty on the emittance measurement.

2 The Electron-Muon Ranger in the MICE Muon Beam

The primary purpose of the EMR is to distinguish between muons and their decay products, identifying muons
that have crossed the entire cooling channel [20]. This allows for the selection of a muon beam with a contam-
ination below 1%. The EMR is a tracking calorimeter, consisting of multiple, orthogonal, layers of triangular
scintillating bars arranged in planes. This construction allows several discriminating parameters to be deter-
mined that may be used to identify particle species at the EMR.

2.1 EMR construction

Figure 2 shows one plane of the EMR, consisting of 59, tessellated, triangular scintillator bars covering an area
of 1.21 m2 . Each plane is rotated through 90◦ with respect to the previous one, such that a pair of planes defines
a horizontal and vertical (x, y) interaction coordinate. Light produced by a particle interaction is collected by

2
Figure 1: (a) Cross-sectional and (b) side views of the MICE Muon Beam Line (MMB) and the MICE ex-
periment. The Electron-Muon Ranger immediately precedes the beam stop and is the final component of the
experiment. The MICE magnet channel (from upstream spectrometer solenoid up to and including the down-
stream sprectrometer solenoid) was not present when the data reported here was taken.

3
 Figure 2: CAD drawing of one EMR plane (top)
and cross section of 3 bars and their threaded WLS
fibres (bottom).
Figure 3: The EMR detector installed in the MICE
Hall at the end of the preliminary (‘Step I’) beam
line. The photograph also shows Q7–9 and TOF1
at the top half of the frame. A model of one spec-
trometer solenoid was positioned in the hall to al-
low installation procedures to be developed.

the wavelength shifting (WLS) fibre within each bar and is read out at both ends. On one side of the plane,
the 59 fibres are bundled together and brought to a conventional single-anode photo-multiplier tube (SAPMT)
and on the other side the fibres are individually brought to one of the pixels on a 64-channel multi-anode
photo-multiplier tube (MAPMT). A dedicated front-end board reads out and digitizes the MAPMT signals
individually by means of fast shapers and discriminators.
Figure 3 shows the full detector, consisting of 48 planes, installed in the MICE Hall at the end of the prelimi-
nary (‘Step I’) beam line [13]. It is enclosed on all sides to ensure a light-tight environment and is magnetically
shielded from the cooling cell. A complete description of the design and fabrication of the EMR can be found
in [20].

2.2 Data taking

In September 2013, the EMR was placed after TOF2 and KL, ∼ 10 m from TOF1, and exposed to a variety
of beams , prior to the installation of the cooling channel. Both ‘muon’ and ‘calibration’ beams over a range
of momenta were used to illuminate the detector. These data were used to verify its ability to separate muons
from their decay products.
The MMB can be tuned to enhance the production of muons (the ‘muon’ beam), pions or electrons (‘cal-
ibration’ beams). All beam configurations contain all three particle species, albeit in differing proportions.
Beams are defined according to the preferred particle species, e, µ, π and the momentum selected (in the range
100–400 MeV/c) by the D2 magnet, pD2 , for the preferred particle type.
Particles travelling from D2 to the EMR cross ∼ 22.7 m of air as well as the remaining PID detectors (TOFs,
Ckov, KL), losing momentum due to ionization. This is expressed as a momentum-loss fraction in figure 4, i.e.
the fraction of momentum lost by a particle given its momentum selected at D2. As the EMR must identify
muons from electrons over a range of momenta, the beam momentum was scanned from 230–450 MeV/c at
D2. This corresponds to ‘muon’ and ‘pion’ beam momenta in the range 200–420 MeV/c, with ‘electron’ beam
momentum range of 100–250 MeV/c.

4
Figure 4: Momentum loss fraction of particles Figure 5: Time-of-flight between the TOF1 and
travelling from D2 to the EMR, compared to their TOF2 detectors (TOF1→2) for a ‘calibration’
momentum at D2 (pD2 ). beam selected with an array of momenta at D2
(pD2 ).

2.3 Particle identification by time-of-flight

The three time-of-flight detectors, TOF0, TOF1 and TOF2, can be used to measure particle momentum, given a
particle-species hypothesis. Calibration beams exhibit a three-peak pattern, corresponding to electrons, muons
and pions in order of increasing time-of-flight as shown in figure 5 for an array of D2 settings. The time-of-
flight for a particular particle is used to define a particle-type hypothesis and to determine the corresponding
momentum via p/E = s/t, where s is the path length between TOFi and TOFj and t = tTOFi − tTOFj .
TOF0 and TOF1 are separated by 7.8 m, whereas TOF1 and TOF2 are separated by 9.4 m. The air between
the last two hodoscopes causes a minimum ionizing muon to lose approximately 2 MeV/c, i.e. ∼ 1% of
its momentum. The calculation of the probability of a particle being classed as an electron, muon or pion
using time-of-flight, for later assessment by the EMR, is based on the assumption that momentum remains
approximately constant and that the path length, s, is given by the orthogonal distance between the two TOF
planes in question. A combined probability from the three TOFs (i.e. TOF0→1, TOF0→2, TOF1→2) of greater
than 90% for a particle to be a muon, leads to that particle being accepted as part of the muon sample, and its
momentum to be calculated from its time-of-flight TOF1→2 using the appropriate particle mass. Similarly, a
combined probability greater than 90% for a particle to be an electron, leads to that particle being accepted as
part of the electron sample. Only the particles that were successfully tagged were used to study the performance
of the EMR.

2.4 Momentum loss in TOF2 and the KL

Before entering the EMR, particles pass through the final TOF station, TOF2, and the pre-shower detector,
KL. The energy lost by particles in TOF2 is calculated using the Continuous Slowing Down Approximation
(CSDA) [21], where the rate of energy loss at every point along its track is assumed to be the same as the
total stopping power. TOF2 is composed of two, 2.54 cm thick, planes of polyvinyltoluene (PVT) constituting
0.1 X0 . On average, minimum ionizing muons lose ∼ 10 MeV/c, whereas electrons lose ∼ 15 MeV/c as they
are ultrarelativistic (βγ > 100) in all cases (figure 6).
The KL is a sampling calorimeter composed of extruded Pb foils and scintillating fibres in a 1 : 2 volume
ratio. On average this corresponds to 2.5 X0 . Electrons shower in the KL before exiting the detector, which
makes estimating the energy lost difficult. Muons cross without showering, but still deposit a significant portion

5
Figure 6: Continuous Slowing Down Approximation of Figure 7: Continuous Slowing Down Approxima-
the mean momentum loss (∆pz ) of particles crossing tion of the mean momentum loss (∆pz ) of muons
TOF2 with impinging momentum pz . crossing the KL with impinging momentum pz .

of their energy in the KL. Due to the complexity of the KL detector, a simplified CSDA approximation was
developed to estimate momentum loss in the KL (figure 7). Minimum ionizing muons lose, on average, 28 ±
3 MeV/c in the KL.

2.5 Events in the EMR

The EMR is put immediately behind the KL. It is an advantage when using it to identify electron-induced
shower since the KL acts as a pre-shower detector.
To characterise the performance of the EMR, particles were first identified with the TOF detectors as de-
scribed in section 2.3. The particle events included in this analysis consist of a single space-point (X- and
Y-view hits) in each TOF and the successful assignment of a particle species. These space points were matched
to the relevant EMR hits through their temporal proximity.
The coordinates of a hit are taken to be those of the centre of gravity of the triangular section of the cor-
responding bar, which is located one third of its height up from its base and bisecting its width. The EMR
planes (defined in section 2.1) are placed perpendicular to the beam and their numbering therefore provides the
z coordinate. The bar number within that plane determines the second coordinate, q, i.e x is the coordinate
provided by the X planes and y the coordinate provided by the Y planes.
An event in the EMR induced by a single positron hitting the KL is represented in the xz and yz projections in
figure 8. The positron originates from an ‘calibration’ beam with a particle momentum of 450 MeV/c selected at
D2. The remaining charged particles in the shower create shallow straight tracks in the first half of the detector,
while photons produce deep and detached hits that are scattered far into the scintillating volume. Some planes
do not record any hit on the particle's path, as might be expected in an electromagnetic shower.
Figure 9 represents the trace left by a single positive muon stopping in the EMR. The impinging muon stops
in the 34th plane where the highest energy deposition is recorded (two bars with a time-over-threshold ∼ 20
ADC counts1 ). The second part of the track, discriminated from the first part using the timing information, is
the positron produced by the muon decay.

1
The time-over-threshold is defined as the width of the low threshold discriminator signal. This signal width, larger for signal with
larger amplitudes, depends on the energy lost by the particle. 1 ADC count corresponds to a signal width of 2.5 ns.

6
Figure 8: EMR event display of the energy deposited by a positron shower (pD2 = 450 MeV/c) in the two
projections. The location of a hit is defined by the plane number (Plane ID, 0–47) and bar number (Bar ID,
1–59) and the energy deposited is represented by the colour code in units of time-over-threshold.

Figure 9: EMR event display of the energy deposited by a µ+ which decays in the detector volume. The
location of a hit is defined by the plane number (Plane ID, 0–47) and bar number (Bar ID, 1–59) and the energy
deposited is represented by the colour code in units of time-over-threshold.

7
3 EMR particle identification variables

Several variables are defined to reduce the information recorded in a complex event to manageable quantities.
Due to the different topology of an electron shower and a muon track, a geometrical approach was chosen.
Hits from an electromagnetic shower are widely spread throughout the detector, including far downstream into
the fiducial volume, without any visible tracks upstream. Muons exhibit clear straight tracks through the EMR
before stopping or exiting the volume.
Because of the photon-induced hits at large depth, the range of an electron event is not well defined and thus
is not a good variable to use to reject electrons. Non-negligible crosstalk [22] within the multi-anode PMTs
prevents the use of a simple shape analysis as muon events could appear wider, thus potentially increasing
the loss of real muons. Alternative longitudinal and transverse evaluations of the event are used to distinguish
between the two species: the ‘plane density’ and the ‘shower spread’, respectively.

3.1 Plane density

This statistic uses the non-continuity of electromagnetic showers to tag electrons and positrons. Electromag-
netic showers in the EMR have multiple and disconnected tracks. Planes on the path of the shower are often
left without any recorded hit. This characteristic can be contrasted with a muon track that uniformly deposits
significant amounts of energy along its path.
Plane density, ρP , is defined as the fraction of planes on the path of the particle that record at least one hit.
More specifically, it is the average of that density in the two projections, i.e:

Nx + Ny N
ρP = = ; (2)
Zx + Zy Zx + Zy
with Nq the number of planes hit in the qz projection and Zq the plane number of the most downstream plane
in the qz projection, q = x, y. N is the total number of planes hit.
The graph in figure 10 shows the plane-density distributions for particles tagged as muons and electrons.
Muons have densities close to 1 while electrons have a widely spread distribution centred around ρP ∼ 0.5. It
is striking to see how few electron entries overlap the muon cases.

3.2 Shower spread

In this discriminator, only the most energetic hit in each plane, i.e the bar that records the largest time-over-
threshold, is kept. This step allows for the rejection of low-energy crosstalk that would dramatically increase
the spread of muon tracks. Crosstalk occurs when some of the light carried by a readout fibre leaks onto
adjacent channels on the MAPMT, causing artificial hits in bars up to 13 cm away from the actual muon track
[22]. Because of the fibre-to-multi-anode PMT mapping, crosstalk only occurs within a single plane, i.e. at a
fixed depth in the detector and never between two layers of the calorimeter.
The remaining Nq hits in a given projection are converted into a set of coordinates (zi , qi ), i = 1, ..., Nq and
a line is fitted through these coordinates; an example is shown in figure 11.
The trajectories of the particles in the two projections may be parametrised using a function of the form
fq (z) = aq z + bq . A linear, least-squares fit was performed by minimizing the normalised χ2 given by:

Nq Nq Nq
1 X (qi − fq (zi ))2 1 X 1 X
χ̂2q ≡ χ2q /νq = 2 = 2 (q i − f (z
q i ))2
= (qi − fq (zi ))2 . (3)
νq σ i ν σ
q 0 i=1 Nq i=1
i=1

8
Figure 10: Logarithmic scale distributions of the plane density (ρP ) for muons and electrons. The integrated
electron sample has been normalised to the number of muons.

Figure 11: Fitted electron shower and muon track in the xz projection. The location of a hit is defined by the
plane number (Plane ID, 0–47) and bar number (Bar ID, 1–59).

9
Figure 12: Logarithmic scale scatter plot of the muon and electron samples in the (χ̂2x , χ̂2y ) space. χ̂2q is the
normalised χ2 of the linear fit in the qz projection with z along the beam line.

PNq −2
with νq = i=1 σi , the normalisation constant. As all the points have the same uncertainty (same bar
geometry), it is possible to extract the common σ02 ≡ σi2 out of the sum which simplifies the development.
Given this definition, it can be shown that the values of aq and bq that minimize the normalised χ2 require that:
• The barycentre (hzi, hqi) of the points belongs to the polynomial, i.e hqi = aq hzi + bq ; and
• aq = Vzq /Vz where Vzq is the covariance of z and q and Vz is the variance of z.
Under these conditions, the normalised χ2 in each projection is:

χ̂2q = σq2 1 − corr2 (z, q) ;

(4)

with corr(z, q) = Vzq /σz σq and σz,q the standard deviation of the z and q coordinates respectively. The
normalised χ2 is proportional to σq2 , i.e the variance of the q coordinate, which means that the normalised χ2
increases with the lateral spread of the event in q.
In figure 12, χ̂2y is plotted against χ̂2x for particles identified as electrons and muons. As expected from the
definition of these quantities, the majority of the muon sample has a small χ2 in each of the two projections
and is located close to the origin. For electrons, the distribution populates the region of large χ̂2x and χ̂2y .

4 Electron-Muon separation

The two variables defined in sections 3.1 and 3.2 were used to develop a general test statistic, T , by which
to distinguish muons from electrons. Given an unknown particle P , boundaries, TC , were defined on these
variables such that the null hypothesis H0 (“is a muon”) or the alternative hypothesis H1 (“is an electron”)
can be tested.
Two parameters were defined to allow for the hypotheses H0 and H1 to be tested. The loss, α, is defined
as the proportion of real muons that test negative to H0 , i.e. an error of the first kind or false negatives. The
contamination, β, is defined as the proportion of real electrons that test positive to H0 , i.e. an error of the

10
Figure 13: Percentage of the electron sample tagged as a muon (β) as a function of the loss of real muons (α)
for different values on the cut ρC . The black dot represents the optimal point of the curve.

second kind or false positives. In this context, the most relevant space to work in is the (α, β) space. A test is
said to be uniformly more efficient if it has a smaller contamination for any given loss [23].
When the choice of test has been made, a value of the optimal cut, TC∗ , has to be calculated. The optimal
value must be such that the corresponding
p point (α∗ , β ∗ ) in the (α, β) space is as close to the origin as possible.
Consider the quantity ∆ = α2 + β 2 as the distance between a point of this space to its origin. The value of
the cut with the smallest distance ∆min is the optimal cut. A perfect test would achieve α∗ = β ∗ = 0, i.e. no
overlap between the two samples.

4.1 Plane density test

Muons tend to have a much higher plane density than their decay products. Given an unknown particle type P ,
define a cut, ρC , such that:
ρP > ρC → H0 ; or ρP ≤ ρC → H1 .

The particle is tagged as a muon if its plane density is greater than ρC and as an electron otherwise. The
muon and electron samples are reprocessed for a wide array of cuts and the values of the contamination and
the loss are computed for each of them. The results of this scanning are represented in figure 13 in the (α, β)
space. The points occupy a single curve in this space because they are the result of a one dimensional cut. The
curve is monotonic as raising the cut increases the loss while reducing the contamination.
The distance, ∆, is calculated for each point and the optimum cut is determined. With this single-variable
cut, more than 99% of the entire muon sample is identified correctly and the electron contamination is ∼2.2%
in the muon sample.

11
Figure 14: Percentage of the electron sample tagged as a muon (β) as a function of the loss of real muons (α)
for three choices of test statistic ξ. The black dot is the optimal point regardless of the choice of test statistic.

4.2 Spread test

The shower spread is summarised by two separate values for the two projections χ̂2x and χ̂2y . The second test
statistic, ξ, is a combination of these two values. The following tests were considered in this analysis:
1. ξ1 = maxq=x,y χ̂2q ; largest of the χ̂2q . A cut on ξ1 limits a square area in the (χ̂2x , χ̂2y ) space;
2. ξ2 = χ̂2x + χ̂2y ; sum of the χ̂2q . A cut on ξ2 limits a triangular area in the (χ̂2x , χ̂2y ) space;
3. ξ3 = χ̂2x × χ̂2y ; product of the χ̂2q . A cut on ξ3 limits a hyperbolic area in the (χ̂2x , χ̂2y ) space.
A study of the electron contamination, β, as a function of the muon loss, α, was performed to determine the
best choice for ξ. The results of this analysis are shown in figure 14. For losses below ∼ 0.5%, the hyperbolic
combination is the most efficient test, but the triangular test yields the lowest contamination for losses above ∼
0.5% and globally gives the best optimal cut. The square cut produces results similar to the triangular one, but
is poorer for low-loss values. Therefore, ξ ≡ ξ2 = χ̂2x + χ̂2y was chosen as the test statistic.
Muons produce very low χ2 straight tracks in the EMR while electrons, that shower in KL, yield much higher
values of χ2 . Given an unknown particle P , define a cut, ξC , such that:

ξ < ξC → H0 ; or ξ ≥ ξC → H1 .

The particle is tagged as a muon if ξ is below a certain threshold and as an electron otherwise. A scan for
different ξC produces the data in figure 14 in the (α, β) space. Once again, for a given choice of test, the points
occupy a single dimension and produce a monotonic curve, when the cut is raised the loss decreases and the
contamination grows.
The optimal point is obtained by minimizing the distance ∆. The spread test is not as efficient as the density
test as both the loss and the contamination are greater. Approximately 97.4% of the muon sample is identified
correctly while the electron contamination of the final sample is ∼ 6.5%. While not optimal in isolation, the
spread test can be used in addition to the density test.

12
Figure 15: Percentage of the electron sample tagged as a muon (β) versus the loss of real muons (α) in the
multivariate analysis. The large black triangle is the optimal point.

4.3 Multivariate test

Combining the two test statistics ρP and ξ yields better results than either statistic alone. Given an unknown
particle species, consider a set of cuts ρC , ξC such that:

ρP > ρC ∩ ξ < ξC → H0 ; or
ρP ≤ ρC ∪ ξ ≥ ξC → H1 . (5)

This is a straightforward combination of the two single-variable tests and represents a triangular prism in the
(ρP , χ̂2x , χ̂2y ) space. It has the advantage of including both features of each particle type into one single test.
Muon events that are both dense and narrow will be tagged H0 while non-continuous and wide electron events
will not.
A scan of both cuts produces the plot in figure 15. The points do not form a single monotonic function in
the (α, β) space, showing that the two variables ρP and ξ are not entirely correlated, i.e. that their combination
carries more information than either one alone. For a given ρC , the contamination increases with ξC .
The most critical element of this electron-muon-separation analysis was to determine the optimal point of this
multivariate test, as it provides the best separation efficiency by combining single-variable tests. The optimum
choice of cuts indicated in figure 15 yields a muon loss of ∼ 0.9% and an electron contamination of below
1.45%. Both figures are lower than the single-variable tests and hence the multivariate analysis is superior to
either single-variable analysis.
For instance, in case of equal abundance of muons and electrons, the EMR will deliver a muon-sample purity
of 98.55% with a loss below 1%. Taking all the beam settings combined, the abundance ratio of electrons to
muons was 11.7%, producing a purity greater than 99.8% in the final muon sample. In the muon-dominated
beam used to demonstrate ionization cooling, the purity provided by the EMR will be much greater than the
99% required by MICE [18].

13
Figure 16: Percentage of electron contamination and muon loss for different ranges of momentum set at D2
(pD2 ). The error bars are based on the statistical uncertainty in a bin.

4.4 Momentum dependence

The previous sections show that the detector functions as designed and reaches the desired efficiency for sam-
ples of muons and electrons. It is necessary to check that the multivariate test developed in this analysis is
consistently powerful for all momenta set at D2.
The optimal cuts, ρ∗C = 0.91 and ξC∗ = 1.6, were applied for different ranges of momentum and the level of

contamination and loss was determined for each range. The loss, α, in figure 16 is ∼1% for the full momentum
range. The contamination, β, fluctuates around 2%.

5 Muon momentum reconstruction

In its cooling demonstration, MICE will measure the phase-space reduction of a muon beam traversing the
cooling cell. This muon beam is selected upstream of the cooling cell by two dipoles in such a way that a
beam with very high muon purity enters the experiment [13]. The two trackers measure the muon momentum
with great accuracy upstream and downstream of the absorber, provided that the muon has a sufficient transverse
momentum, pT . It is not important for the EMR to be able to measure the pion momentum or that of the rejected
electrons. It is interesting, however, to estimate how well the EMR can reconstruct the muon momentum. The
EMR can assist the trackers with low pT muon tracks and can provide a redundant measurement for larger pT .

5.1 Range reconstruction

In order to avoid problems induced by MAPMT crosstalk, the most energetic hit in each plane, selected in the
same way as was done for the estimation of the spread, was used for the range reconstruction.

14
Figure 17: Muon range as a function of the momentum reconstructed from TOF1→2. At 280 MeV/c and above,
muons can traverse the entire detector without stopping, hence the plateau at R ∼ 816 mm.

The reconstruction method uses the linear fit defined in section 3.2. The muon track was fitted in its two
separate projections and the pitch of its trajectory was measured in each of them. The angles in the xz and
yz projections are θx and θy , respectively. Provided that the last bar hit by the track has the coordinates
(xN , yN , zN ), the total range is:
q
R = zN 1 + (tan2 θx + tan2 θy ) ≡ zN / cos θ ; (6)
p 
with θ = tan−1 tan2 θx + tan2 θy , the total angle with respect to the z axis. The results are shown for the
entire muon sample in figure 17.

5.2 Theoretical approximation

In order to give an estimate of the muon momentum from a single range measurement it is necessary to develop
a theoretical understanding of the dependence of range on momentum. A fit of the data recorded is used to
provide a single estimate of the muon momentum downstream of the cooling channel, pd , as a function of the
range in the EMR. We define an invertible function, f , such that:

R = f (pd ) ⇐⇒ pd = f −1 (R) . (7)

To construct a manageable function, the EMR geometry is reduced to a simple 816 mm thick block of
polystyrene as it is the base material used in the plastic scintillator bars. Finding the range of muons in the
EMR is equivalent to computing the range in polystyrene. In this situation, the best approach is the same CSDA
method used to calculate the energy loss. For an average muon stopping power, hdE/dxi, of polystyrene, the
CSDA range is: Z 0 Z 0
dE dp
R= = βmµ c2 . (8)
E0 hdE/dxi p0 /mµ c hdE/dxi

15
Figure 18: Muon range as a function of the initial momentum p0 = pd − ∆p(pd ). The error bars represent the
bin width in abscissa and the uncertainty of the average reconstructed range in ordinate.

The integral in eq. (8) has no analytical solution due to the complexity of the Bethe-Bloch formula. How-
ever, it can be integrated numerically for a muon of initial momentum p0 . The function can also be inverted
numerically to give an estimate of the momentum.
pd is the momentum of the muon just before it enters TOF2 and corresponds to the momentum reconstructed
from TOF1→2. This constitutes a complication because it is significantly different from the initial momentum
in the EMR, p0 . The function f −1 provides an estimate of the range for a given p0 , but must compensate for
the combined momentum loss in TOF2 and KL, ∆p(pd ). The correction, p0 = pd − ∆p(pd ), has to be applied
muon by muon.
Figure 18 shows the range of muons in the EMR as a function of their initial momentum, p0 . The CSDA
prediction shows very good agreement with data for p0 > 150 MeV/c while the agreement is acceptable for
lower p0 (the plateau of the simulation is added to fit the dimensions of the EMR). The agreement could
be improved by a more detailed analysis of the KL, this is beyond the scope of this paper. However, the
approximation presented here is a powerful tool that allows an estimate of momentum to be made.

5.3 Momentum reconstruction accuracy

Provided with an estimation of the initial momentum, p0 , the uncertainty on the unfolding of pd may be esti-
mated. There are several sources of uncertainty:
• The uncertainty on the time of flight σt ∼ 70 ps and path length σs ∼ 1 cm;
• Fluctuation of the momentum-loss in KL as a function of pd ;
• The uncertainty on the longitudinal position in the EMR, σz ∼ 5 mm.
Figure 19 shows that the RMS of the momentum distribution in a given plane in the centre of the detector is
∼ 10 MeV/c. Each bin represents a single plane of the EMR (17 mm in depth). At the entrance of the fiducial
volume, muons of low momentum stop in the first few planes and the RMS is large. At the back of the detector,

16
Figure 19: Downstream momentum (pd ) as a function of the range (R) in the EMR. A bin corresponds to a
plane and its error bar represents the RMS of the momentum distribution within the bin.

the muons with momentum above 320 MeV/c cross the entire detector and thus the last two planes are biased
towards higher values.
The uncertainty on pd emerging from the TOF momentum reconstruction, with t the time-of-flight measure-
ment, can be shown to be:
 2 !−3/2  2 !−3/2
mµ c2 t ct mµ c2 t2 ct
σTOF = −1 σt ⊕ −1 σs , (9)
s2 s s3 s
with ⊕ indicating the quadratic sum. σTOF ranges from 3–15 MeV/c for momenta from 150–350 MeV/c. As a
bin contains a range of momenta, eq. (9) is averaged over a normal p.d.f. centred on the average momentum in
the bin and with a width equal to the bin RMS in order to produce σTOF .
The spread of the energy loss in KL is computed using the distribution of momentum loss, fp (∆p), given for
an array of momenta, p, as shown in figure 7. A single momentum-loss distribution, fi (∆p), constructed for a
bin centred in Ri , is defined as:
Z +∞ (p−p )2
1 − σ2i
fi (∆p) = fp (∆p) √ e i dp ; (10)
0 πσi
2
(p−pi )
√ −
σ2
with 1/( πσi )e i the normal distribution centred on the average momentum in the bin, pi , and with
width equal to the bin RMS, σi . The RMS of fi (∆p) is used as the uncertainty on the momentum from the
momentum loss in KL, σKL .
The total uncertainty on the momentum from the range measurement, Rµ , can be summarized as:
σR = σTOF ⊕ σKL ⊕ σEMR ; (11)
and therefore, given that the uncertainties are not correlated:
q
σEMR ∼ σR 2 − σ2 2
TOF − σKL ; (12)

17
 Figure 19 shows a compilation of the uncertainties originating from the different components of this analysis.
For small values of the range, the EMR has a poor resolution and does not produce reliable momentum recon-
struction. After about 10 planes, the total spread reaches its nominal value of ∼ 10 MeV/c and is constant until
the last two planes where the values are biased by the muons exiting the detector. The EMR itself, subtracting
the errors from KL and TOF, achieves resolutions down to 3 MeV/c for larger range.

6 Conclusions

The ability of the Electron-Muon Ranger (EMR) to separate electrons from muons has been demonstrated. It is
known from simulation [24] that muons and electrons produce distinct signal patterns: muons below a certain
energy stop in the detector and the energy deposition exhibits a clear Bragg peak, the position of which defines
the range. Electrons shower in KL and exhibit multiple tracks in the EMR. In order to verify this, the EMR was
exposed to a beam containing both types of particles with momenta in the range 100–400 MeV/c. As expected,
electrons and muons produce substantially different patterns so that electrons can be identified with 98.6%
efficiency using the EMR alone.
The detector was not tuned or optimized prior to the measurements. Nevertheless, the detector was able
to separate electrons from muons using a simple test statistic on two variables based on the transverse and
longitudinal structure of an event.
The detector provides tracking and calorimetric information. Tracks and showers can be reconstructed and
identified as muons or electrons. A Bragg peak at the end of muon and pion tracks marks the place where a
particle stops and, therefore, helps in measuring the range. It was shown that the muon range can be used to
infer the particle momentum in the Continuously Slowing Down Approximation with very good accuracy.

18
Acknowledgements

The work described here was made possible by grants from Department of Energy and National Science Foun-
dation (USA), the Instituto Nazionale di Fisica Nucleare (Italy), the Science and Technology Facilities Council
(UK), the European Community under the European Commission Framework Programme 7 (AIDA project,
grant agreement no. 262025, TIARA project, grant agreement no. 261905, and EuCARD), the Japan Society
for the Promotion of Science and the Swiss National Science Foundation, in the framework of the SCOPES
programme. We gratefully acknowledge all sources of support.
We are also grateful to the staff of the ISIS Department at the Rutherford Appleton Laboratory for the reliable
operation of ISIS. We acknowledge the use of Grid computing resources deployed and operated by GridPP in
the UK, http://www.gridpp.ac.uk/.
This publication is dedicated to the memory of Mike Zisman who passed away while the data presented here
was being prepared for publication.

19
References

[1] IDS-NF Collaboration, S. Choubey and et al., “International Design Study for the Neutrino Factory,
Interim Design Report,” tech. rep., 2011. arXiv:1112.2853.

[2] M. Alsharo’a and et al. Phys. Rev. ST Accel. Beams 6 (Aug, 2003) 081001.

[3] V. Parkhomchuck and A. Skrinsky, “Cooling Methods for Charged Particle Beams,” Rev.Accel.Sci.Tech.
1 no. 1, (2008) 237.

[4] ISS Accelerator Working Group Collaboration, M. Apollonio and et al., “Accelerator design concept
for future neutrino facilities,” Journal of Instrumentation 4 no. 7, (2009) .

[5] R. Palmer, A. Sessler, A. Skrinsky, A. Tollestrup, A. Baltz, et al., “Muon collider design,”
Nucl.Phys.Proc.Suppl. 51A (1996) 61–84, arXiv:acc-phys/9604001 [acc-phys].

[6] MICE Collaboration, G. Gregoire, G. Ryckewaert, L. Chevalier, J. Rey, M. Catanesi, et al., “Proposal to
the Rutherford Appleton Laboratory: an International Muon Ionization Cooling Experiment (MICE),”.
http://mice.iit.edu/micenotes/public/pdf/MICE0021/MICE0021.pdf.
MICE-NOTE-21.

[7] D. Neuffer, “Principles and Applications of Muon Cooling,” Part.Accel. 14 (1983) 75.

[8] D. Neuffer, “Multi-TeV Muon Colliders,” AIP Conf.Proc. 156 (1987) 201–208.

[9] M. Palmer and et al., “Muon accelerators for the next generation of high energy physics experiments,”
13th International Particle Accelerator Conference (IPAC 13), Shanghai, China, 2013.

[10] “Muon ionization cooling experiment,”. http://mice.iit.edu.

[11] R. Asfandiyarov, “A totally active scintillator calorimeter for the Muon Ionization Cooling Experiment
(MICE). Design and construction,” Nucl.Instrum.Meth. A732 (2013) 451–456.

[12] R. C. Fernow and R. B. Palmer, “Solenoidal ionization cooling lattices,” Phys. Rev. ST Accel. Beams 10
(Jun, 2007) 064001.

[13] MICE Collaboration, M. Bogomilov and et al., “The MICE Muon Beam on ISIS and the beam-line
instrumentation of the Muon Ionization Cooling Experiment,” Journal of Instrumentation 7 no. 5, (2012)
P05009.

[14] MICE Collaboration, V. Blackmore, C. Hunt, J. Pasternak, C. Rogers, P. Snopok, and H. Witte, “MICE
Ionization Cooling Demonstration: Technical Note,”.
http://mice.iit.edu/mnp/MICE0452.pdf. MICE-NOTE-DET-452.

[15] “Isis pulsed neutron and muon source at the rutherford appleton laboratory,”.
http://www.isis.stfc.ac.uk.

[16] C. N. Booth and et al., “The design, construction and performance of the MICE target,” Journal of
Instrumentation 8 no. 03, (2013) P03006.

[17] R. Bertoni and et al., “The design and commissioning of the MICE upstream time-of-flight system,”
Nucl. Inst. Meth. A 615 no. 1, (2010) 14 – 26.

20
[18] MICE Collaboration, M. Rayner and M. Bonesini, “The MICE PID Detector System,” tech. rep., 2010.
http://mice.iit.edu/mnp/MICE0304.pdf. MICE-NOTE-304.

[19] MICE Collaboration, R. Bertoni, A. Bonesini, A. de Bari, G. Cecchet, Y. Karadzhov, and R. Mazza,
“The construction of the MICE TOF2 detector,”. http://mice.iit.edu/mnp/MICE0286.pdf.
MICE-NOTE-DET-286.

[20] R. Asfandiyarov, Totally Active Scintillator Tracker-Calorimeter for the Muon Ionization Cooling
Experiment. PhD thesis, University of Geneva, Jul, 2014. CERN-THESIS-2014-164.

[21] N. Carron, An Introduction to the Passage of Energetic Particles through Matter. Taylor & Francis,
2006. https://books.google.co.uk/books?id=LRAHeoBP_l8C.

[22] MICE Collaboration, R. Asfandiyarov, A. Blondel, and F. Drielsma, “Crosstalk and Misalignment in the
Electron Muon Ranger (EMR),”. http://mice.iit.edu/mnp/MICE0440.pdf.
MICE-NOTE-DET-440.

[23] F. James, Statistical Methods in Experimental Physics. World Scientific Publishing, 2006.
https://books.google.co.uk/books?id=S8N_QgAACAAJ.

[24] MICE Collaboration, R. Asfandiyarov, A. Blondel, and L. Haegel, “Geant4 Simulation of the EMR
Detector Response Study,”. http://mice.iit.edu/mnp/MICE0388.pdf.
MICE-NOTE-DET-388.

21
The MICE collaboration

M. Bogomilov, R. Tsenov, G. Vankova-Kirilova

Department of Atomic Physics, St. Kliment Ohridski University of Sofia, Sofia, Bulgaria

R. Bertoni, M. Bonesini, F. Chignoli, R. Mazza

Sezione INFN Milano Bicocca, Dipartimento di Fisica G. Occhialini, Milano, Italy

V. Palladino
Sezione INFN Napoli and Dipartimento di Fisica, Università Federico II, Complesso Universitario di Monte
S. Angelo, Napoli, Italy

A. de Bari, G. Cecchet
Sezione INFN Pavia and Dipartimento di Fisica, Pavia, Italy

M. Capponi, A. Iaciofano, D. Orestano, F. Pastorea , L. Tortora

Sezione INFN Roma Tre e Dipartimento di Fisica, Roma, Italy

Y. Kuno, H. Sakamoto
Osaka University, Graduate School of Science, Department of Physics, Toyonaka, Osaka, Japan

S. Ishimoto
High Energy Accelerator Research Organization (KEK), Institute of Particle and Nuclear Studies, Tsukuba,
Ibaraki, Japan

F. Filthautb
Nikhef, Amsterdam, The Netherlands

O. M. Hansen, S. Ramberger, M. Vretenar

CERN, Geneva, Switzerland

R. Asfandiyarov, P. Bene, A. Blondel, F. Cadoux, S. Debieux, F. Drielsma∗ , J. S. Graulich, C. Husi, Y. Karadzhov,

F. Masciocchi, L. Nicola, E. Noah Messomo, K. Rothenfusser, R.Sandström, H. Wisting
DPNC, Section de Physique, Université de Genève, Geneva, Switzerland

G. Charnley, N. Collomb, A. Gallagher, A. Grant, S. Griffiths, T. Hartnett, B. Martlew, A. Moss, A. Muir,

I. Mullacrane, A. Oates, P. Owens, G. Stokes, P. Warburton, C. White
STFC Daresbury Laboratory, Daresbury, Cheshire, UK

D. Adams, P. Barclay, V. Bayliss, T. W. Bradshaw, M. Courthold, V. Francis, L. Fry, T. Hayler, M. Hills, A. Lin-
tern, C. Macwaters, A. Nichols, R. Preece, S. Ricciardi, C. Rogers, T. Stanley, J. Tarrant, S. Watson, A. Wilson
STFC Rutherford Appleton Laboratory, Harwell Oxford, Didcot, UK

a
Deceased
b
Also at Radboud University, Nijmegen, The Netherlands
∗
Corresponding author.

22
R. Bayes, J. C. Nugent, F. J. P. Soler
School of Physics and Astronomy, Kelvin Building, The University of Glasgow, Glasgow, UK

P. Cooke, R. Gamet
Department of Physics, University of Liverpool, Liverpool, UK

A. Alekou, M. Apollonio, G. Barber, D. Colling, A. Dobbs, P. Dornan, C. Hunt, J-B. Lagrange, K. Long,
J. Martyniak, S. Middleton, J. Pasternak, E. Santos, T. Savidge, M. A. Uchida
Department of Physics, Blackett Laboratory, Imperial College London, London, UK

V. J. Blackmorec ,T. Carlisle, J. H. Cobb, W. Lau, M. A. Rayner, C. D. Tunnell

Department of Physics, University of Oxford, Denys Wilkinson Building, Oxford, UK

C. N. Booth, P. Hodgson, J. Langlands, R. Nicholson, E. Overton, M. Robinson, P. J. Smith

Department of Physics and Astronomy, University of Sheffield, Sheffield, UK

A. Dick, K. Ronald, D. Speirs, C. G. Whyte, A. Young

Department of Physics, University of Strathclyde, Glasgow, UK

S. Boyd, P. Franchini, J. Greis, C. Pidcott, I. Taylor

Department of Physics, University of Warwick, Coventry, UK

R. Gardener, P. Kyberd, M. Littlefield, J. J. Nebrensky

Brunel University, Uxbridge, UK

A. D. Bross, T. Fitzpatricka , M. Leonova, A. Moretti, D. Neuffer, M. Popovic, P. Rubinov, R. Rucinski

Fermilab, Batavia, IL, USA

T. J. Roberts
Muons, Inc., Batavia, IL, USA

D. Bowring, A. DeMello, S. Gourlay, D. Li, S. Prestemon, S. Virostek, M. Zismana

Lawrence Berkeley National Laboratory, Berkeley, CA, USA

P. Hanlet, G. Kafka, D. M. Kaplan, D. Rajaram, P. Snopok, Y. Torun

Illinois Institute of Technology, Chicago, IL, USA

S. Blot, Y. K. Kim
Enrico Fermi Institute, University of Chicago, Chicago, IL, USA

U. Bravar
University of New Hampshire, Durham, NH, USA

Y. Onel
Department of Physics and Astronomy, University of Iowa, Iowa City, IA, USA
c
Now at Department of Physics, Blackett Laboratory, Imperial College London, London, UK

23
 L. M. Cremaldi, T. L. Hart, T. Luo, D. A. Sanders, D. J. Summers
University of Mississippi, Oxford, MS, USA

D. Clinea , X. Yang
University of California, Los Angeles, CA, USA

L. Coney, G. G. Hanson, C. Heidt

University of California, Riverside, CA, USA

24

View publication stats